Protect pulled data in LatencyAggregator with a lock
The data pull happens on a binder thread, which is different from the
thread where the sketches are collected.
Before this CL, the data was modified by the puller, so it could
potentially be corrupted if the user happened to use the device at the
same time.
This is a speculative fix to the infinite loop that we are observing
inside CompactStack.
Bug: 298423577
Test: atest libinput_tests inputflinger_tests
(cherry picked from https://googleplex-android-review.googlesource.com/q/commit:ad16bf27ab695919347965d917de3c1c7ed5cb0f)
Merged-In: I624fb3bd59e3c314edc3b1facf424c306e95b71e
Change-Id: I624fb3bd59e3c314edc3b1facf424c306e95b71e
diff --git a/services/inputflinger/dispatcher/LatencyAggregator.cpp b/services/inputflinger/dispatcher/LatencyAggregator.cpp
index 96d78c3..e09d97a 100644
--- a/services/inputflinger/dispatcher/LatencyAggregator.cpp
+++ b/services/inputflinger/dispatcher/LatencyAggregator.cpp
@@ -126,6 +126,7 @@
}
void LatencyAggregator::processStatistics(const InputEventTimeline& timeline) {
+ std::scoped_lock lock(mLock);
// Before we do any processing, check that we have not yet exceeded MAX_SIZE
if (mNumSketchEventsProcessed >= MAX_EVENTS_FOR_STATISTICS) {
return;
@@ -167,6 +168,7 @@
}
AStatsManager_PullAtomCallbackReturn LatencyAggregator::pullData(AStatsEventList* data) {
+ std::scoped_lock lock(mLock);
std::array<std::unique_ptr<SafeBytesField>, SketchIndex::SIZE> serializedDownData;
std::array<std::unique_ptr<SafeBytesField>, SketchIndex::SIZE> serializedMoveData;
for (size_t i = 0; i < SketchIndex::SIZE; i++) {
@@ -257,6 +259,7 @@
}
std::string LatencyAggregator::dump(const char* prefix) const {
+ std::scoped_lock lock(mLock);
std::string sketchDump = StringPrintf("%s Sketches:\n", prefix);
for (size_t i = 0; i < SketchIndex::SIZE; i++) {
const int64_t numDown = mDownSketches[i]->num_values();
diff --git a/services/inputflinger/dispatcher/LatencyAggregator.h b/services/inputflinger/dispatcher/LatencyAggregator.h
index 60b6813..d6d1686 100644
--- a/services/inputflinger/dispatcher/LatencyAggregator.h
+++ b/services/inputflinger/dispatcher/LatencyAggregator.h
@@ -16,6 +16,7 @@
#pragma once
+#include <android-base/thread_annotations.h>
#include <kll.h>
#include <statslog.h>
#include <utils/Timers.h>
@@ -61,10 +62,13 @@
~LatencyAggregator();
private:
+ // Binder call -- called on a binder thread. This is different from the thread where the rest of
+ // the public API is called.
static AStatsManager_PullAtomCallbackReturn pullAtomCallback(int32_t atom_tag,
AStatsEventList* data,
void* cookie);
AStatsManager_PullAtomCallbackReturn pullData(AStatsEventList* data);
+
// ---------- Slow event handling ----------
void processSlowEvent(const InputEventTimeline& timeline);
nsecs_t mLastSlowEventTime = 0;
@@ -74,14 +78,17 @@
size_t mNumEventsSinceLastSlowEventReport = 0;
// ---------- Statistics handling ----------
+ // Statistics is pulled rather than pushed. It's pulled on a binder thread, and therefore will
+ // be accessed by two different threads. The lock is needed to protect the pulled data.
+ mutable std::mutex mLock;
void processStatistics(const InputEventTimeline& timeline);
// Sketches
std::array<std::unique_ptr<dist_proc::aggregation::KllQuantile>, SketchIndex::SIZE>
- mDownSketches;
+ mDownSketches GUARDED_BY(mLock);
std::array<std::unique_ptr<dist_proc::aggregation::KllQuantile>, SketchIndex::SIZE>
- mMoveSketches;
+ mMoveSketches GUARDED_BY(mLock);
// How many events have been processed so far
- size_t mNumSketchEventsProcessed = 0;
+ size_t mNumSketchEventsProcessed GUARDED_BY(mLock) = 0;
};
} // namespace android::inputdispatcher